Subject: Re: Processor correctavke error?
To: Matthew Jacob <mjacob@feral.com>
From: Chris G. Demetriou <cgd@pa.dec.com>
List: port-alpha
Date: 06/10/1998 11:43:06
> I'm working on trying to expand the amount of usable information
> reported in these kinds of traps. This one is a system found
> error- despite what Chris and others say, I still think the
> pc and ra can be interesting - what's the pc and ra correspond
> to for this kernel, and what've you got loaded as options on
> this box?

I think the PC and RA can be interesting, but may not necessarily be
in the case of the problems you're looking at.  (In general, I've
found over my experience with the port that machine checks or other
faults that reference locations in locore and other lowest-level fault
handling code are often indicators of unrelated problems, that happen
to trigger asynchronously in those places...  That's how I look at
them.  It's perfectly fine that you have your own take on how they
should be looked at, however.)

In the output Ken posted:

>unexpected machine check
>
>    mces    = 0x1
>    vector  = 0x670
>    param   = 0xfffffc0000006048
>    pc      = 0x12000b71c
>    ra      = 0x12000b964
>    curproc = 0xfffffe004a575000
>        pid = 56, comm = gzip

The PC and RA are in the (or at least a) user-land process.  (Maybe
you were expecting a PC and RA stored elsewhere, but those are the
ones that he's likely to see w/o additional code, right?)



cgd